Simon Michael <simon@joyful.com>, 2002/06/01 01:23 GMT (via mail):
Overhauled the creation_time and last_edit_time fields: they are now both ISO-format strings, and creationTime & lastEditTime methods give DateTime? versions. I hope this is similar to the way CMF etc. does it - anyone know ?. Updated this sites' *Changes pages and standard_wiki_header accordingly.

Rewrote the instance-upgrading code, now called upgrade (and upgradeSubscribers). It now enforces the ISO-format time strings, does not change last_editor when upgrading, and should be more reliable particularly with subscriber lists (perhaps solving IssueNo0161?).

Lost all the last_editors on this site (I break stuff so you don't have to..). last_edit_time is now displayed by preference (over bobobase_modification_time) everywhere except the tracker, which has yet to be converted. One of the side benefits has been more accurate results in RecentChangesZC? etc.

I have catalog indexes & metadata in place for last_edit_time & creation_time, and also lastEditTime which I needed for FastChanges?, and the new links method, which lists all links in a page.


Simon Michael <simon@joyful.com>, 2002/06/02 01:44 GMT (via mail):
There has been an expansion and grand renaming of page types: see QuickReference and AllAboutPageTypes?.

All seem to be pretty solid and the ZwikiPerformance figures show that pre-linking pays off, so that's close to being made the default. I have switched all the zwiki.org pages to pre-linking modes (and pre-rendered every page) so things should be pretty quick around here.

All 16 page types & variants are available in the edit form at the moment but I will reduce this to a subset. In fact this whole site could probably get by quite well with only stxprelinkdtmlhtml now (SearchPage? & RecentChanges? used to require htmldtml, but no longer).


Simon, 2002/06/02 03:30 GMT (via web):
Out of memory again. 3-day pack: 225Mb -> 130Mb


"Jay, Dylan" <djay@avaya.com>, 2002/06/02 23:52 GMT (via mail):

> Overhauled the creation_time and last_edit_time fields: they

Now that edit times are no longer linked to the OFS perhaps we can think about the other big painful link, ie History. Packing losses the history which is useful info. I think Zope.org solves this problem by using seperate mounted ZODB. Anyone have any other solutions or thoughts on the "best" solution to this problem?

On a similar note, another history based feature I'd like to see (that I think will be very useful), is CVS style annotation. This is a view of the current page where every line has a name and time by it showing who was the last person to edit that line (I realize html is not line based so some smarts might have to be used). This gives a good one page look at where the major content has come from and which bits are new vs old. I think this would be very valuable in collaborative environment. Perhaps if keeping the whole history is expensive/hard, just keeping enough information for this kind of annotated view would be enough? Thoguhts?


"Jay, Dylan" <djay@avaya.com>, 2002/06/03 07:12 GMT (via mail):
Under previous versions of ZWiki I built a FAQ system that enabled people to create questions that get transformed into wiki links such as [How do I get help with this system?]?

The problem was that ? chars are disallowed in Zope object ids. I got around this by hacking zope which is not an ok solution. With the latest version of Zope this isn't even enough as ZWiki won't recognize the above as a valid link.

My current thinking on how to support this is to hack ZWiki (or better comeup with a general solution to supporting strange chars). This solution would entail changing

My question is, has anyone tried to do anything like this already? Does anyone have any advice about the best approach? Is anyone else interested in this kind of functionality?


Simon Michael <simon@joyful.com>, 2002/06/03 19:17 GMT (via mail):
"Jay, Dylan" <djay@avaya.com> writes:
> which is useful info. I think Zope.org solves this problem by using seperate
> mounted ZODB. Anyone have any other solutions or thoughts on the "best"
> solution to this problem?

Swiki does this well - simple, stable numbered revisions. To achieve this, we might

  1. use some other kind of zodb storage - not an option for most people
  2. store our own edit history. The simplest way I can think of implementing this would be for each page to save all it's past texts in an attribute. Might cause unacceptable zodb bloat with large pages/large wikis ?
  3. store our own edit history, but use the file system. We should explore fruitful uses for filesystem storage with zwiki.


> think will be very useful), is CVS style annotation. This is a view of the
> current page where every line has a name and time by it showing who was the
> last person to edit that line (I realize html is not line based so some
> smarts might have to be used). This gives a good one page look at where the

Yes this would be very cool.


Simon Michael <simon@joyful.com>, 2002/06/03 20:08 GMT (via mail):
"Jay, Dylan" <djay@avaya.com> writes:
> Under previous versions of ZWiki I built a FAQ system that enabled people to
> create questions that get transformed into wiki links such as
> [How do I get help with this system?]?

I was just doing some FAQ work - but all in a single page. (cf CreatingFAQsInZwiki). Yes it would be nice to make per-question pages with a name like that.


> The problem was that ? chars are disallowed in Zope object ids. I got
> around this by hacking zope which is not an ok solution. With the latest
> version of Zope this isn't even enough as ZWiki won't recognize the
> above as a valid link.

The same problem applies with international characters. I made zwiki follow the zope standard on this so that it wouldn't lead to an error when you tried to create the link. It's hard-coded in Regexps.py if you want to hack it (it should probably import the regexp so that you only have to hack it once).


> My question is, has anyone tried to do anything like this already? Does
> anyone have any advice about the best approach? Is anyone else interested in
> this kind of functionality?

Yes I am.

Standard wiki links enforce a simple naming principle that's part of wiki's success, ie link text == page name == url. At present, zwiki's bracketed links follow that principle. Except, spaces etc. are necessarily url-quoted; so even though some browsers will accept spaces, in fact the true url is different from the link text.

We could let bracketed links work differently. In particular, we could say "bracketed page names may be anything" and when someone creates a page from one of these links, we will assign a url-safe (probably different) id for it. The page name will go in the hitherto unused title property, and links will be based on that.

In addition, something else I was thinking about: when comparing bracketed links with existing page names, ignore whitespace, punctuation (?), and case. Ie it will link if you have the correct sequence of letters & digits.

What do you think ?


"Jay, Dylan" <djay@avaya.com>, 2002/06/03 23:28 GMT (via mail):

> -----Original Message-----
> From: Simon Michael
> Sent: Tuesday, 4 June 2002 5:18 AM
> To: @joyful.com
> Subject: zwiki.org GeneralDiscussion
>
>
> "Jay, Dylan" <djay@avaya.com> writes:
> > which is useful info. I think Zope.org solves this problem by using
> seperate
> > mounted ZODB. Anyone have any other solutions or thoughts on the
> "best"
> > solution to this problem?
>
> Swiki does this well - simple, stable numbered revisions. To
> achieve this,
> we might
>
> 1. use some other kind of zodb storage - not an option for most people
>
> 2. store our own edit history. The simplest way I can think of
> implementing this would be for each page to save all it's
> past texts in an
> attribute. Might cause unacceptable zodb bloat with large pages/large
> wikis ?
>
> 3. store our own edit history, but use the file system. We
> should explore
> fruitful uses for filesystem storage with zwiki.

  1. Just store the annotations of who wrote which line when. Not much bloat then. Not useful for going backward I've never found that too useful with regard to wikis (beyond stort-term, between packs fixing mistakes). Could be a good short term solution until more flexibility in storage is available.


> > think will be very useful), is CVS style annotation. This is a view
> of the
> > current page where every line has a name and time by it showing who
> was the
> > last person to edit that line (I realize html is not line based so
> some
> > smarts might have to be used). This gives a good one page look at
> where the
>
> Yes this would be very cool.
>
> --
> forwarded from http://zwiki.org/GeneralDiscussion
>


"Jay, Dylan" <djay@avaya.com>, 2002/06/03 23:45 GMT (via mail):

> From: Simon Michael
>
> "Jay, Dylan" <djay@avaya.com> writes:
> > Under previous versions of ZWiki I built a FAQ system that enabled
> people to
> > create questions that get transformed into wiki links such as
> > [How do I get help with this system?]?
>
> I was just doing some FAQ work - but all in a single page. (cf
> CreatingFAQsInZwiki). Yes it would be nice to make
> per-question pages with
> a name like that.

Mine worked by having an add question form at the footer. This would append the question as inside [ ]?. Worked great except for question marks.


> > The problem was that ? chars are disallowed in Zope object ids. I
> got
> > around this by hacking zope which is not an ok solution. With the
> latest
> > version of Zope this isn't even enough as ZWiki won't recognize
> the
> > above as a valid link.
>
> The same problem applies with international characters. I made zwiki
> follow the zope standard on this so that it wouldn't lead to
> an error when
> you tried to create the link. It's hard-coded in Regexps.py
> if you want to
> hack it (it should probably import the regexp so that you only have to
> hack it once).
>
> > My question is, has anyone tried to do anything like this already?
> Does
> > anyone have any advice about the best approach? Is anyone else
> interested in
> > this kind of functionality?
>
> Yes I am.
>
> Standard wiki links enforce a simple naming principle that's part of
> wiki's success, ie link text == page name == url. At present, zwiki's
> bracketed links follow that principle. Except, spaces etc.
> are necessarily
> url-quoted; so even though some browsers will accept spaces,
> in fact the
> true url is different from the link text.
>
> We could let bracketed links work differently. In particular, we could
> say "bracketed page names may be anything" and when someone

I like this. This makes life much easier for users.


> creates a page
> from one of these links, we will assign a url-safe (probably
> different) id
> for it. The page name will go in the hitherto unused title
> property, and
> links will be based on that.
>
> In addition, something else I was thinking about: when
> comparing bracketed
> links with existing page names, ignore whitespace,
> punctuation (?), and
> case. Ie it will link if you have the correct sequence of letters &
> digits.

This would be might be a problem since "A Question" would be the same as "aquestion" and "A, question?". Maybe that's not a problem, maybe that's a benefit? Since the you're be removing the less meaninful stuff it would allow people to link to what they mean much easier.

I wouldn't remove all whitespace however. I'd trim it. Otherwise you would be changing the meaning.


> What do you think ?

Since your doing quoting for spaces why not just do url quoting all the way? The quoted name gets stored... crap! I just realized that Zope doesn't allow % either!!! That means unless we hack Zope your solution is the only way to go.

I'm not sure what the problem the zope guys have with international chars and punctuation. Since they allowed spaces in it's no longer guarenteed python compatible, so why don't they allow full url quoting?? or more chars? If it's allowed in a url it should be allowed in a objectManaged name. (My2c)


Simon, 2002/06/04 00:55 GMT (via web):
Well a few delicate and not-so-delicate hacks later and.. it seems to be working on this site. The "bracketed names may be anything" part, not yet the fuzzy matching part. Check it out! TestPage


"Jay, Dylan" <djay@avaya.com>, 2002/06/04 01:27 GMT (via mail):

> Well a few delicate and not-so-delicate hacks later and..
> it seems to be working on this site. The "bracketed names
> may be anything" part, not yet the fuzzy matching part.
> Check it out! TestPage

Looking good. What's the chance of having a downloadable version ready for when I upgrade my FAQ Zope in 5.5 hours time?


Simon, 2002/06/04 01:48 GMT (via web):
And there's the fuzzy matching. You can update from CVS now, or you can download the nightly cvs tarball (FrontPage, top right) when it gets rebuilt - scheduled at midnight PST, so you might just be ok.

I'm liking this - hope nothing bad shows up to bite us. :)


"Jay, Dylan" <djay@avaya.com>, 2002/06/04 01:55 GMT (via mail):

> download the nightly cvs tarball (FrontPage, top right) when
> it gets rebuilt -
> scheduled at midnight PST, so you might just be ok.

Now all I need is a script to convert the id's of my old pages to what they will be called now. You don't have the conversion routine handy do you so I could start on that now?


> I'm liking this - hope nothing bad shows up to bite us. :)

As long as all the old chars are still handled as they used to be.... what about spaces, arn't they removed now?


Simon, 2002/06/04 02:09 GMT (via web):
You don't have zeo or python access to your zodb, do you ? That makes it easy. I'm not up to date with pythonscripts. In dtml, something like this might get you part of the way:

 <dtml-in "wikifolder.objectValues(spec='ZWiki TestPage')">
 <dtml-let 
  title=id
  title="_.string.replace(title,'A',' A')"
  title="_.string.replace(title,'B',' B')"
  etc...
 >
 <dtml-call "manage_changeProperty('title',title) #or whatever is it">
 </dtml-in>

or in emacs, I'd open the folder via ftp and rename the files with a keyboard macro.


2002/06/04 02:10 GMT (via web):

>> I was just doing some FAQ work - but all in a single page. (cf CreatingFAQsInZwiki).

I see you noticed my many attempts at updating the faq page. :( I ultimately ended up with HTML. Thanks for the follow-up -Dean


Mark Chambers, 2002/06/04 02:15 GMT (via web):
Regarding spaces in names - why not replace them with an underscore rather than drop them. This makes uses_pace and use_space clearly different and still easy to read.


Simon, 2002/06/04 02:19 GMT (via web):
All old pages should work as before. Pages created after this upgrade will have

  1. an id - always zope id-safe, url-safe, and a valid wikiname (the zwiki variant)
  2. optionally: a more readable title - anything you like, will be displayed in preference to the id

Links can be formed either by writing the bare id/wikiname, or by writing the title in [].


Simon, 2002/06/04 02:22 GMT (via web):
[uses pace]? (UsesPace?) is different from [use space]? (UseSpace?)


Simon, 2002/06/04 02:30 GMT (via web):
Dean I noticed you added a bunch of stuff - good work!


simon, 2002/06/04 21:53 GMT (via web):
WikiWikiWeb:EvolutionOfaWikizen


Simon Michael <simon@joyful.com>, 2002/06/04 23:58 GMT (via mail):
at last, we do something useful when you enter a non-existent url. It will look up pages by title, or do a jumpsearch. (I made this work as a quick fix for title links in the parent context display).

Can't tell if it's slow, or everything is today.. the site is not loaded.

[standard_error_message/view_source]?


2002/06/05 09:07 GMT (via web):
Is there going to be a restructuredtext formatting for pages? It would be great, esp. for more "formal" postings, e.g. copy-pasting some internal README files and such.

-- Ville


Simon Michael <simon@joyful.com>, 2002/06/05 17:22 GMT (via mail):
That's up to you all.. I'm all for it :)


"Jay, Dylan" <djay@avaya.com>, 2002/06/06 05:32 GMT (new page):
AnyChar? created


"Jay, Dylan" <djay@avaya.com>, 2002/06/06 05:38 GMT (via mail):
.... That didn't work as I intended. Not sure I like that kind of email functionality. Here is what I meant to say

------------------------

Simon, I thin you need to update the edit function arount line 1355 in ZWiki.py. It's test to check if a page exists or not doesn't take into account of the new name conversion.

Can you also make public the routines to check if a name exists or not?

Thanks to your very nice work the customization needed for my FAQ almost comes down to just this python script

comment = "\n - [" + question + "]\n"
context.FrontPage.comment(text=comment, use_heading=1,
REQUEST=container.REQUEST)

id = context.FrontPage._pageWithFuzzyTitle(self, question) if not hasattr(context.FrontPage.aq_parent, page): # set up default content specific to a question page text = "This is a NewFAQQuestion which has yet to be answered}\n\n" text = text + "To answer just Edit this page and replace this text with your answer" context.FrontPage.edit(text=text, page=question)

which currently won't work due to a presky underscore.


rob, 2002/06/06 14:58 GMT (via web):
Hi, What file exactly would I modify if I didn't want colspan to appear in tables, e.g.

see source for td colspan=bla

yuk

bla

bla

bla

It's not wwml.py ... ANS: it's in StructuredText/HTMLClass?.py under lib/python on Zope...


Simon Michael <simon@joyful.com>, 2002/06/06 16:21 GMT (via mail):
Dylan - made them public yesterday. Is that what you needed ?

These aren't finalised yet. It may be that having eg [a page with !]? and [a page with ?]? (or [ÃTestPage]? and [APage]?) link to the same thing will be too unintuitive ?

If we can live with it, we can keep our simple rule for generating the wikiname from any title. If not, then we have to come up with another rule for mapping these characters into a url-safe name. Ideally, the safe name needs to be easy to guess, read, type, and not clash with any other legal name.


simon, 2002/06/06 16:52 GMT (via web):
[Ã TestPage]? and TestPage, rather. NB these links don't work right when included on FrontPage yet, look at GeneralDiscussion instead.


Simon Michael <simon@joyful.com>, 2002/06/06 17:57 GMT (via mail):
By the way, I'd like to solicit any ideas and actions relating to the following. It feels like it's needed.

Quality Issues ( ZwikiQuality )

...


2002/06/06 19:27 GMT (via web):
I'm interested in the WikiMail concept. I took the liberty of adding a link to WikiMail to refer to the WikiMailDiscussion. Some questions:

/file.doc/200206060002 - returns the second file.doc uploaded today

/file.doc - returns current version of file.doc

/GeneralDiscussion/200206060001 - First message of today

I'll try to spend some time to working on this feature in my holidays.

PieterB (I tried sending this message using e-mail, using the subject Status of WikiMail, but it doesn't seemed to work)


2002/06/06 19:41 GMT (via web):
I've added a WikiSubscriber? page where everybody can introduce themselves a bit.

PieterB


pieterb@gewis.nl (Pieter Biemond (prive)), 2002/06/06 20:12 GMT (via mail):
Ok, I found out where my "Status of WikiMail" message to zwiki@zwiki.org went. It was appended to WikiMail. This was not what I had expected. It might be a good idea to use "mailto:zwiki@zwiki.org?subject=GeneralDiscussion" instead of the current mailto-link.

I think the bottom of the FrontPage can be improved, by rewriting it to something like:

more GeneralDiscussion
Feel free to add your comments via the web.
You can also e-mail comments to zwiki@zwiki.org if you are subscribed to one or more wiki's.

I'm wondering what happens if the users has more than one WikiName in the subject. This message has the subject "GeneralDiscussion WikiMail". Is/should this message be appended to both wiki's or only to the first mentioned in the subject?

PieterB


Simon, 2002/06/06 21:22 GMT (via web):
Hi Pieter,

with multiple wikinames in the subject, only the first one is used. Re FrontPage, you can edit it of course. I was thinking of adding a special comment form there which would post to the GeneralDiscussion page.


Simon, 2002/06/07 00:20 GMT (via web):
Switched to a new page id generating algorithm. This one still attempts to generate a wikiname that is as close as possible to the free-form title, but does not strip international characters or punctuation. It uses _ for quoting in urls, like MoinMoin. See TestPage for some examples, here's the docstring:

    def pageIdFromTitle(self,title):
        """
        convert a page title made up of arbitrary text to a canonical safe id

        Constraints for zwiki page ids:
        - it needs to be a legal zope object id
        - to simplify linking, we will require it to be a valid url
        - it should be unique for a given title (ignoring whitespace)
        - we'd like it be as similar to the title and as simple to read
          and work with as possible

        So this version
        - strips leading/trailing whitespace
        - capitalizes and joins the whitespace-separated words into a wikiname
        - converts any non-zope-and-url-safe characters (and _) to _hexvalue
        - if the last step results in an id beginning with _, prepend X


Simon, 2002/06/07 02:19 GMT (via web):
Disabled the standard_error_message for now, until I can teach it not to gobble all errors. Fixed a buglet with the free-form titles that was breaking GeneralDiscussion.


EdwardKreis, 2002/06/07 07:53 GMT (via web):
There seems to be the bug in ZWikiPage.stxToHtml method - it doesn't take into account, that Zope adds not only html and body tags but also title tag in between. For example you can see the source for the TestPage3?, there are seen additional html, body tags after the ZWiki header.

The last lines rewriten as here does the work.:

        # strip html & body if added (not all zope versions support header=0)
        text = re.sub(
            r'(?sm)^<html>\n(.*)<body>\n(.*)</body>\n</html>\n',r'\2',text)
        return text


Simon, 2002/06/07 14:55 GMT (via web):
Thanks, fixed.


CraegStrong, 2002/06/07 20:49 GMT (via web):
What about the possibility of allowing the contents of Wiki pages to be stored in the filesystem, rather than in the ZODB? That way they could be easily stored in a revision control system, etc. and all types of file-based tools could be used without having to use FTP or what-not. Possibly only a small fraction of users would actually want this feature, but for the few, this would be extremely nice. In fact, we needed this feature for our Zope so badly that I actually wrote a product to provide it (for certain Zope object types). The technology is very straightforward and could be readily adapted to ZWiki pages. Thoughts?


simon, 2002/06/08 04:56 GMT (via web):
1-day pack. 300Mb -> 55Mb. Things are getting desperate. :)


AlV, 2002/06/08 15:59 GMT (via web):
I adhere to the local fs proposal.


simon, 2002/06/08 18:20 GMT (via web):
Craeg, welcome. I think it's well worth exploring.

More on the zodb: I am working through the implementation issues with free-form titles and fuzzy linking. Both are currently enabled here, and backlinks is using the new linkids method which should allow accurate backlinks alongside fuzzy linking. This stuff is expensive; I wasn't able to index linkids without running out of memory so I ran this script at the zeo prompt:

 >>> for p in w.objectValues(spec='ZWiki TestPage'):
 ...   print p.id()
 ...   try:
 ...     p.reindex_object()
 ...   except:
 ...     print 'FAILED'
 ...   get_transaction().commit()
 ...   app._p_jar.sync()
 ... 

This indexed the thousand pages successfully, but also grew the zodb from 50 to 250Mb. :) I don't know if linkids is the cause or if my catalog in general is just too big and inefficient. Currently I index and store metadata for everything except the page text.


simon, 2002/06/08 20:17 GMT (via web):
ZODB/catalog experiments (everything but page text is indexed & metadata'd):

  0-day pack: 250 -> 17Mb
  removed linkids index & metadata
  updated catalog: zodb grew 17Mb -> 23Mb
  observed zope memory size at 100Mb.. left over from before ?
  restarted zope, updated catalog again: memory reached 50Mb, zodb grew 5Mb
  updated catalog again: zodb grew another 5Mb.. it's like that eh
  reindexed just the links field: zodb grew .4Mb
  re-added the linkids index, reindexed it: 
   memory grew to 100Mb, browser timed out as zope worked for 5-10 minutes
   (expected since linkids is slow), apparently succeeded, zodb grew .5Mb
  re-added linkids to metadata, reindexed linkids again:
   looked like above but taking much, much longer - killed after one hour;
   no significant zodb growth, several conflict errors, site unusable
   during this time

So rebuilding the whole linkids index is expensive in cpu time and memory; updating the whole catalog is expensive in zodb bloat. That makes sense. I don't understand what's happening with the last step.


simon, 2002/06/08 20:49 GMT (via web):
here we go one more time.. bear with me..


simon, 2002/06/08 21:35 GMT (via web):
I don't actually need metadata for linkids in order to do accurate (that is, fuzzy) backlinks, just the index. However calculating backlinks in this way implies that when we create or delete a page, we must look up all it's backlinks and reindex them. Hmm, and how do we look them up exactly ? Chicken and egg problem. Reindexing all pages would work. Brrr.

Accurate backlinks are cool when your catalog is all current; for the moment I have reverted to the usual exact-matching backlinks here.


simon, 2002/06/08 22:31 GMT (via web):
And this may not be a bad moment to mull over the WikiWikiWeb:WikiDesignPrinciples


DeanGoodmanson, 2002/06/09 16:37 GMT (via web):
I've venturing into CMF land, and CMFWiki is taking some getting used to. Particularly, I was pulling my hair out over trying to mask out Wiki name recognition with the ! character. It does seem to be a bit more STX focused than my current habits.

Suggestions for resources for learning navigating & managing CMF also appreciated, via email, DeanGoodmanson, or here if appropriate.


"Jay, Dylan" <djay@avaya.com>, 2002/06/11 01:08 GMT (via mail):

> Dylan - made them public yesterday. Is that what you needed ?

Yes, seems to be working well


> These aren't finalised yet. It may be that having eg [a page
> with !] and
> [a page with ?]? (or [=C3Page]? and [APage]?) link to the same
> thing will be
> too unintuitive ?

I think that is fair enough. I think most search engines do this kind of translation for instance. But perhaps someone who is working with an i18n wiki would be better placed to comment. All I needed was the question mark :)


"Jay, Dylan" <djay@avaya.com>, 2002/06/11 01:20 GMT (via mail):

> - how to improve people's typical first encounter with zwiki ?

Probably make it look slicker I guess. Seems to impress people.


> - how to increase awareness of known bugs & solutions in
> current releases ?

How about, CurrentBugs? and ChangesLog? type pages being released as default pages in a release. These would be linked to the FrontPage and have links to the last info on zwiki.org


> - how to increase awareness and ease of upgrading to latest
> code & template ?

Upgrading is still a pain. Every upgrade I have to find and delete the default pages and headers/footers and then copy over the new set from a newly created Zwiki folder. There has to be a better way than this. Perhaps the way CMF does it? Perhaps someway to merge wiki pages during install? Pehaps refactoring to make common customizations exist in files that shouldn't be replaced. One way to do this for instance might be with ZPT and macros. A standard_wiki_view.zpt could be customizable, but it could reference macros to do "functionality" type stuff.


> - how to improve people's experiences at zwiki.org ?

Knowing how old info is I think is helpful for any wiki and something zwiki doesn't handle 100% well. I think I saw this one on another wiki. Use tool tip on links to say how old the page it links to is (could just use the ALT tag to do this), and who last edited the page.


2002/06/11 01:39 GMT (via web):
Simon, - how to improve people's typical first encounter with zwiki ? and also how to improve people's experiences at zwiki.org? First, the user will probably say to himself, what's in it for me using the zwiki.org
Interface more attractive, emphasize on main links for better navigation. Not to overwhelm the first user with lot of text on the Front page, just to the point links. For instance, I was trying to post comments, it is true that English is not my primary language, but was wondering how to find the add a comment box to post my comment.
-Jo


"Jay, Dylan" <djay@avaya.com>, 2002/06/11 02:06 GMT (via mail):

> Simon,
> - how to improve people's typical first encounter with
> zwiki ? and also how
> to improve people's experiences at zwiki.org?

> First, the user will probably say to himself, what's in it
> for me using the
> zwiki.org

> Interface more attractive, emphasize on main links for better
> navigation. Not
> to overwhelm the first user
> with lot of text on the Front page, just to the point links.
> For instance, I
> was trying to post comments, it is true that English is not my primary
> language, but was wondering how to find the add a comment box
> to post my
> comment.

A comment on a similar theme would be to turn all cababilites on by default and let people turn them off. Uploading images is a function that most people don't know about because they have to turn it on.


2002/06/11 02:51 GMT (via web):
Daydreaming about WikiMail enhancements...

I would like each comment tagged with an < a name="CCYYMMDDHHMM"> tag, and that tag listed in the email sent out : e.g forwarded from: http://zwiki.org/GeneralDiscussion#200206110939

I'll add to to this one after posting your for experimentation.

This may make it easier for people to reference a discussion point they're replying to, also. (Then again, that may get cryptic, let's hope that ole fashion correspondence protocol remains.)

Many other nits, including archiving. :( Food for disection, I guess..


"Jay, Dylan" <djay@avaya.com>, 2002/06/11 03:33 GMT (via mail):

>
> Daydreaming about WikiMail enhancements...

WikiLinks? being turned into real links in the email might be good. This might mean the use of HTML mail which might be bad :( Maybe just an ugly link in brackets next to the WikiName? or perhaps an email appendix with a name -> link index?


2002/06/11 13:53 GMT (via web):
I wouldn't ues GeneralDiscusion?#20020606 but instead link to GeneralDiscussion/20020606. Then make the GeneralDiscussion automagically list all the messages from the last month (and links to previous months). PieterB


JohnGreenaway, 2002/06/11 17:07 GMT (via web):
Regarding tooltips on links - I really like this. The code below (for v0.9.9) does it nicley. The "title" attribute on an href link is supported by most fairly recent browsers. The code replaces the same chunk in the _renderlink() method (and _replaceWikilink(), not sure which is used now...) in wikipage.py:

        # a wikiname - if a page (or something) of this name exists in
        # this folder, link to it
        elif hasattr(self.aq_inner.aq_parent.aq_base, m): 
        #elif m in self.aq_inner.aq_parent.objectIds():
            #return '<a href="%s">%s</a>' % (quote(m), m)
            # wiki links were made absolute for robustness in editform etc.
            # a relative_urls property can re-enable the old behaviour
            linked_page = getattr(self.aq_parent,m)
            if getattr(self.aq_parent,'relative_urls',None):
                return '<a href="%s" title="last edited by %s on %s">%s</a>' % \
                (quote(m),getattr(linked_page,'last_editor'),getattr(linked_page,'last_edit_time'), m)
            else:
                return '<a href="%s/%s" title="last edited by %s on %s">%s</a>' % \ 
                (self.wiki_url(),quote(m),getattr(linked_page,'last_editor'), \
                getattr(linked_page,'last_edit_time'), m)


DeanGoodmanson, 2002/06/11 22:04 GMT (via web):
Jay, Dylan & PieterB: My suggestion was merely a convenience to be directed to the message being refered to in the WikiMail.. HTML would be annoying/bad. Links are good. Email readers that turn URL's to links are handy.

PieterB: I'm not making the connection to the / seperator for the discussion page & the post. Sounds mystical.


>> WikiWikiWeb:EvolutionOfaWikizen - My "Favorites" menu resembles that! (By the way, I suggest using your personal Wiki instead of Favorites experiment.)

John: Thanks for the title attr...I'd been wanting that piece of info for awhile.


"Jay, Dylan" <djay@avaya.com>, 2002/06/11 23:30 GMT (via mail):

> >> WikiWikiWeb:EvolutionOfaWikizen - My "Favorites" menu resembles
> that! (By the way, I suggest using your personal Wiki instead
> of Favorites
> experiment.)

Which segues into another thought I had last night. I too use a personal wiki for favorites. It does lack the convinience however of a one or two click "add to favorites" button. Anyone experimented with a javascript link that can be embeded in an IE toolbar (no idea about other browsers) that would automatically add the link and the title to a certain page? Avantgo has something like this... in fact here it is "javascript:location.href=http://ui.avantgo.com/mydevice/autoadd.html?url= +escape(window.location)+'&title='+escape(document.title);" I should have a go at massaging that into a wiki comment url. But what about quickly adding to a group? I wonder if there is any possible webdav/IE favourite type intergration....


"Jay, Dylan" <djay@avaya.com>, 2002/06/12 00:55 GMT (via mail):
Is there anyway to hide my email address as a result of sending a comment via WikiMail? I get enough spam as it is.


2002/06/12 02:53 GMT (via web):

>> Is there anyway to hide my email address as a result of sending a comment via WikiMail? I get enough spam as it is.

Excuse me while I experiment with the following short-term end-user possibilities:

a. Don't fill in a UserName? in the UserOptions? - BUT... that will require you to sign your posts...and other stuff (this comment posted without a UserName? )

b. Remove your email address in the UserOptions? . BUT... this will come back when you resubscribe, and you'll have to go remove it again.


DeanGoodmanson, 2002/06/12 02:55 GMT (via web):
This post posted with no email address in UserOptions?.

And, by the way, with the TimeZone? of CST ... anyone know if that's the right one for Central? Right now it's 9:44 and my previous post listed 2:53 GMT.


DeanGoodmanson, 2002/06/12 03:09 GMT (via web):
...and to finish things up:

Dylan: Cool idea on the IE toolbar thing. Mozilla has it's own extensions, sidebar's I thnk they're refered too, and revered by many. (Mark Hammond built a cool Python docs search tool.)


>> But what about quickly adding to a group?

a. I'd suggest the "Bookmarklet" Javascript prompt the user for a quick comment on the site, and put it into a general "sort me later" bin , this way it is less likely to get forgotten.

b. Speaking of prompts, could it prompt you with a (predefined) selection list of topics (eg. Wiki Pages) to add the link to?

Barely getting a handle on HTML ... now Javascript is asking for attention! ;-) Three cheers for ZWiki (and I suppose ZoPE?) as the "shell environment" of the web development world!

Off topic, I'm had a few URL not get completely recognized by the ZWiki URL recognizer. Is this known or should I submit these URL's to a bug report? (Crazily enough I was reading about regular expressions and found one for recognizing URLs?, ecxept it included a checkmark character...am I seeing things? I'm at home, the link is at home. I'll post it per reader interest.)

Finally, don't forget to nominate your favorite unsung hero in the Python, Perl and TCL world at the ActiveState? Active Awards 2002 http://www.activestate.com/Corporate/Awards/ActiveAwards.html


Simon, 2002/06/12 18:23 GMT (via web):
Travelling to ireland - back online in a few days. Found a cheap way to do fuzzy backlinks, enabled on this site. Good ideas above!


JohnGreenaway, 2002/06/13 09:12 GMT (via web):
Another great idea from Jay. And here we go... This works in both IE and Mozilla.

To use:


JohnGreenaway, 2002/06/13 09:16 GMT (via web):
Argh. Wiki is replacing the single quotes in the above link with "code" tags. How do I stop it doing that? Anyway, view the source of the page to see the proper link.


JohnGreenaway, 2002/06/13 09:31 GMT (via web):
OK. The "Add to wiki" link above is fixed. Replaced the single quotes in it with the html control char for it. Seems to stop wiki changing them to code tags...


DeanGoodmanson, 2002/06/13 21:07 GMT (via web):
"isn't there a c implementation of the re module ? I don't think this server is using it ?" - Simon , GeneralDiscussion200205

"Regular expression patterns are compiled into a series of bytecodes, which are then executed by a matching engine written in C." - "Regular Expression HOWTO" , http://py-howto.sourceforge.net/regex/node2.html ( http://www.python.org/doc/howto/. )

This seems to answer the question, but maybe not to the extent of your question.


2002/06/16 20:25 GMT (via web):
I am trying out a Zwiki at home. How do I configure it to send mail to subscribers?


2002/06/17 12:08 GMT (via web):
Have a look at the WikiMail page -- PieterB


DeanGoodmanson, 2002/06/17 23:22 GMT (via web):
Has anyone configured they're Wiki with an ExternalEditor link from the "edit" link?

http://www.zope.org/Members/Caseman/ExternalEditor/

Seems Pie-in-the-sky, but breaking free of the text box page editor is a dream of mine...at least for the search/replace capabilities. I understand they're a way to integrate it with a popular text editor...but psuedo WYSIWYG would be take it to the next level...or potentially make the Wiki pages completely unreadable.


Simon Michael <simon@joyful.com>, 2002/06/18 10:03 GMT (via mail):
deltab-mj, your email address is unreachable today. This broke mail-in so I have unsubscribed you for now.


2002/06/18 14:27 GMT (via web):
Hmm, starting to work with ZWiki CVS on Zope 2.6.0a1, I get the following error when adding a "ZWiki page" from the ZMI:

Error Type: AttributeError?
Error Value: aq_inner

I can't find a HTML-traceback in the source (this might be a new security feature op Zope 2.6).

Any suggestions of how to debug this weird behaviour?

PieterB


Simon Michael <simon@joyful.com>, 2002/06/18 15:11 GMT (via mail):
we have a broken tracker add form at present.. will fix when I get some time.


2002/06/18 15:36 GMT (via web):
More WikiMail/SPAM discussion:
I have add a fix for http://www.zwiki.org/IssueNo0066#bottom (spam protection). Can you change that in the CVS?
I think it would be great if users could provide a RealName? when they subscribe to the wiki. That name should have precendence over the realname supplied in the user's e-mail. What do you think? -- PieterB


Simon, 2002/06/18 17:01 GMT (via web):
Tracker fixed (it was IssueNo0159? again)


Simon, 2002/06/18 17:31 GMT (via web):
Thanks to Juan David for the first independent cvs checkin! (spanish translations of the first batch of internationalised strings), and to Joanne Plouffe (canadian french translations).


2002/06/19 10:58 GMT (via web):
Is there a way to go from viewing a page to editing the page without using the mouse?


JohnGreenaway, 2002/06/19 12:47 GMT (via web):

>Is there a way to go from viewing a page to editing the page without using the mouse?

You could just shift-tab a couple of times to get up to the edit link. Or standard_wiki_footer could be changed so the edit link is:

 <a href="&dtml-wiki_page_url;/editform" ACCESSKEY="E">edit this page</a>

Then on a page you could just go alt-e (then enter in IE, no need in Mozilla) to change it.


2002/06/19 13:33 GMT (via web):
Thanks, that's cool, now [alt-e]? opens for edit, [tab enter]? saves and returns to view. One Thing; after [alt-e]? opens the edit window, the textarea doesn't have focus, I need to click on it. Any hints? I'm using Mozilla 1.0 Thanks, Kent


JohnGreenaway, 2002/06/19 14:36 GMT (via web):
You could change the body tag in editform to:

 <body bgcolor="#FFFFFF" onload="document.forms[0].text.focus();">

Javascript, but won't cause any problems to those with js turned off. Just helps the majority with it. Google does the same focus trick when you visit their page.


"Jay, Dylan" <djay@avaya.com>, 2002/06/19 23:16 GMT (via mail):

> Is there a way to go from viewing a page to editing the page
> without using the
> mouse?

If only html supported keyboard shortcuts. Might be possible via javascript I guess.


"Jay, Dylan" <djay@avaya.com>, 2002/06/19 23:19 GMT (via mail):

> >Is there a way to go from viewing a page to editing the page without
> using the mouse?
>
> You could just shift-tab a couple of times to get up to the
> edit link. Or
> standard_wiki_footer could be edited so that the link is::
>
> edit
> this page

Wow, you learn something new (about html) everyday.


2002/06/25 05:30 GMT (via web):
Is there a central respository that details all/most/some/any companies/individuals who run ZWiki sites? Here's my idea: There is way too much lateral replication of information (data with meaning) within the WWW. It's pervasive and pernicious.
If you think about it, the lateral replication of information is contributory to the deaths of people the world over. How do I come to this conclusion, for all of you snickering, well simply this. The unneeded, extra hours of electricity usage involved within the lateral replication of information comes about at a price, an ecologocial price of air pollution, which is contributory toward lung cancer, e.g.,
What do I propose? Shared efforts, e.g., one site take upon the responsibility of bibliographies, one site to handle tech definitions, etc. Interested parties, contact me at pierrejohnson@ureach.com.
Eventually, I plan to create the UURI, the universal unversal resource identifier. There should be one central repository that accurately lists a domain, and its content matter. This kind of activity should not be within the domain of greedy companies and their ridiculous idea: UDDI. That's just a money grubbing bunch of corporations duping the little guy.


2002/06/25 05:39 GMT (via web):
ZWiki Behavior: When creating a hypertext link via "something here" , the "http://reference", one cannot place a colon within the "something here" part. We track book titles within a bibliography. Often, book titles have sub-titles separated by a colon.


2002/06/25 05:41 GMT (via web):
I'd like to see a parsing management page that allows the admin to change the characters parsed by ZWiki.


"Jay, Dylan" <djay@avaya.com>, 2002/06/25 05:50 GMT (via mail):

> -----Original Message-----
> From: zwiki@zwiki.org
>
> I'd like to see a parsing management page that allows the
> admin to change the
> characters parsed by ZWiki.

As in what the regular expression for what a WikiName is? I'd like to see that two. In my wiki's I quite like TLA's to be WikiName but this isn't the default.


Simon Michael, 2002/06/26 09:49 GMT (via mail):

> Is there a central respository that details all/most/some/any
> companies/individuals who run ZWiki sites?

Yes, it's either ZWikiSites? or PeopleUsingZwiki.


Simon Michael, 2002/06/26 09:51 GMT (via mail):
zwiki@zwiki.org writes:
> Hmm, starting to work with ZWiki CVS on Zope 2.6.0a1, I get the following
> error when adding a "ZWiki page" from the ZMI:
>
> Error Type: AttributeError?

> Error Value: aq_inner

>
> I can't find a HTML-traceback in the source (this might be a new security
> feature op Zope 2.6).
>
> Any suggestions of how to debug this weird behaviour?

Thanks for this Pieter.. I added it to the tracker today. Using zope 2.5.0 I did get a traceback. I couldn't copy it at the time (was using w3m browser with mouse support turned on) but I posted the essentials on the issue page.


Simon, 2002/06/26 11:08 GMT (via web):
YES! Stomped that darn IssueNo0161?


2002/06/26 17:33 GMT (via web):
I've noticed that freezope and I think the NIP free zope services are using version 0.8? .

I would like to suggest an upgrade, but not sure which version I should suggest?

I'd like a version with the mass-subscription annoyance fix ( IssueNo0161? ) ... Does the current source have a unique version #?


Simon, 2002/06/27 08:39 GMT (via web):
I'm reluctant to recommend cvs (circa 20020626) as a stable upgrade for general users because it has some significant changes (pre-rendering, pre-linking, fuzzy linking etc) which haven't yet been documented in release notes. (If you are familiar with these changes, I do recommend it).

It would be nice if we had a stable branch in cvs (last release + important bugfixes) but so far this has seemed like too much work. We are getting closer to stabilising 0.9.10. Meanwhile my recommendation for cautious upgraders would be to install 0.9.9 and check KnownIssues for anything important to you, and patch those manually.


Simon, 2002/06/27 12:56 GMT (via web):
7-day zodb pack, 460 -> 228Mb.

The fix for IssueNo0194? should be handy - no more need to escape wikinames inside <a></a>


Simon, 2002/06/27 23:12 GMT (via web):
Enabled ExternalEditor support and JohnGreenaway's neat access key, editform focus and link title suggestions here on zwiki.org. Most header/footer links and most wikilinks should have meaningful link titles now. I tried "last edited by %s on %s" but I think I prefer the current more swiki-like "last edited % ago by %s". Mouse over a wiki link to see it.

As a side issue, what would you think of using the latter format in the page header and possibly even in RecentChanges? ? It has the advantage of being timezone-independent.


Simon, 2002/06/27 23:19 GMT (via web):
..whereas the first format is independent of the time at which this page was rendered.

I forgot to link the list of AccessKeys.

Be gentle while using external editor. Once we figure out the glitches and gotchas it should be a big help.


Jay, Dylan, 2002/06/27 23:31 GMT (via mail):

> As a side issue, what would you think of using the latter
> format in the
> page header and possibly even in RecentChanges? ? It has the
> advantage of being
> timezone-independent.

Since u never want to cache a Wiki then I think it's an excellent idea. I mean that's what people really want to know, how old it is.


Simon Michael, 2002/06/28 00:25 GMT (via mail):
On TextFormattingSurprises, someone wrote words to the effect of:
> I upgraded from an older version of zwiki recently (was like 0.6.4,,now
> at 0.9.9 & it broke a lot of my pages where I'd used pre tags to format
> text as fixed width - the old behavior was that the html tag worked, and
> WikiWords? were links. The new behavior is that text containted between
> pre tags are non-wikified.
>
> Do I have any options to do fixed-width plain text that has wiki
> formatting?

Good question.. I tried using a style attribute on TestPage but didn't get it working. Maybe someone on G.D can help.


JohnGreenaway, 2002/06/28 12:27 GMT (via web):
Trying out the various page markups. Stxdtmllinkhtml seems fine, but with stxprelinkdtmlhtml each page view seems to be creating a new copy of the page in the "history" management screen. This bloats zodb and make the page diffs give lots of blanks. Anyone else get this? No big problem as we can use stxdtmllinkhtml instead though...

Also really like all those new tooltips. Jacob would be proud. And the relative edit times at the top. And the AccessKeys.


Simon Michael, 2002/06/28 14:42 GMT (via mail):
zwiki@zwiki.org (JohnGreenaway) writes:
> with stxprelinkdtmlhtml each page view seems to be creating a new copy
> of the page in the "history" management screen.

That's good to know.. I thought something was. Here's a description of the linking code after this morning's refactoring, for those interested:

    # XXX refactoring in progress. A couple of ways to get stuff linked:
    #
    # 1. the old way
    # _oldRenderLinksIn(text) or wikilink(text) - called at view time
    #  replaces interwiki links in text
    #   calls _replaceInterWikilink for each
    #  replaces other links in text
    #   calls _replaceWikilink for each
    #
    # 2.1 the new way (link arbitrary text)
    # renderLinksIn(text) - called at view time
    #  replaces links in text
    #   calls _renderLink for each
    #
    # 2.2 the new way (link this page's text with caching)
    # _preLink() - called at edit time
    #  searches for valid links in self._prerendered
    #  caches them with text regions in self._prelinked (also self._wikilinks)
    # _renderWithLinks() - called at view time
    #  calls _updateLinkRenderings()
    #   calls _renderLink for each cached link
    #    calls _replaceInterWikilink & emulates _replaceWikilink
    #   caches rendered link in self._wikilinks
    #  interpolates rendered links with cached text regions
    #  saves final text back in self._prerendered, re-cooking if it changed

_renderWithLinks and it's helpers are probably to blame, perhaps because of the way they save the latest link renderings on each view. It should be possible to avoid saving a new version if no link renderings have changed (ie none of this page's link targets have been added or deleted since last rendering).


Simon, 2002/06/28 15:05 GMT (via web):
Ie TestPage._wikilinks looks like:

 {
  'InternationalCharacterExamples': '<a href="/InternationalCharacterExamples" title="last edited 3 weeks ago by simon">InternationalCharacterExamples</a>', 
  '[\xc3 link containing international characters]': '<a href="/X_c3LinkContainingInternationalCharacters" title="last edited 3 weeks ago by Simon">\xc3 link containing international characters</a>',
  etc.

As you can see including the last edit interval in link titles interferes with link caching somewhat. The link rendering will change more frequently especially with young link targets, eg "last edited 30 seconds ago".. "last edited 2 minutes ago".. etc. Other ideas welcome.


Simon Michael, 2002/06/29 09:45 GMT (via mail):
contents currently takes about 1.5 minutes - I'll fix it soon. Also I have some changes to check in which should fix the excessive db writes by prelink modes.


Simon Michael, 2002/06/29 16:43 GMT (via mail):
Freeform links and link titles should now be working in most places on zwiki.org, such as: included text processed by wikilink(); backlinks; recent/fastchanges; parent context; contents; zwikitracker/filterissues. Some links (written as html) may still need tweaking, eg SearchPage?. Post if you find more places.

Pre-linking pages no longer generate a zodb write every time you view them. A write-on-view should now happen only if one of the link targets has been added or deleted since last viewing. Because I like having dynamic link titles, I now render them in a final pass separate from link caching & dtml.

A potential downside of showing current target information in link titles: on each page view every link target is woken up, which might have a memory/speed impact when there are many links. Eg RecentChanges?, contents.

Also, to get the link title for freeform names in contents/parent context I would need to look up the true id. That's expensive so I just punt for now.


Simon, 2002/06/29 17:20 GMT (via web):
I'm looking for examples or ideas for detecting autoresponder replies (out of office, etc) in the mailin script..


Jay, Dylan, 2002/06/30 23:20 GMT (via mail):


> As you can see including the last edit interval in link
> titles interferes with
> link caching somewhat. The link rendering will change more frequently
> especially with young link targets, eg "last edited 30
> seconds ago".. "last
> edited 2 minutes ago".. etc. Other ideas welcome.

Don't pre-render the page. Perhaps cache some expensive stuff like working out what's a link or not, but not the actual page rendering.


Jay, Dylan, 2002/06/30 23:36 GMT (via mail):

> A potential downside of showing current target information in
> link titles:
> on each page view every link target is woken up, which might have a
> memory/speed impact when there are many links. Eg
> RecentChanges?, contents.

Use a ZCatalog. Put the last modified time as metadata. That's one of the main reasons to use a catalog, to cache data in one place so you don't have to wake up lots of objects.


Jay, Dylan, 2002/06/30 23:38 GMT (via mail):

> I'm looking for examples or ideas for detecting autoresponder
> replies (out of
> office, etc) in the mailin script..

Don't solve a solved problem. Integrate with mailman. Look how big their auto-responder code is, why do you want to reinvent that? I think a tight integration between wiki and listserver would be a hot feature.